Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
|
@@ -1,16 +1,9 @@
|
|
| 1 |
-
# Version: Corrected After Test 4 (V2.4.1 - Fixed Validation & Timestamps Button Restored)
|
| 2 |
-
# Description: Cette version corrige la validation des segments modifiés manuellement.
|
| 3 |
-
# Réintégration du bouton "Générer les timestamps".
|
| 4 |
-
# Correction du bug "too many values to unpack".
|
| 5 |
-
# La génération du fichier ZIP fonctionne correctement après validation.
|
| 6 |
-
|
| 7 |
import os
|
| 8 |
import shutil
|
| 9 |
import zipfile
|
| 10 |
import torch
|
| 11 |
-
import numpy as np
|
| 12 |
-
from pathlib import Path
|
| 13 |
import gradio as gr
|
|
|
|
| 14 |
from pydub import AudioSegment
|
| 15 |
from transformers import pipeline
|
| 16 |
|
|
@@ -41,7 +34,7 @@ def init_metadata_state():
|
|
| 41 |
def transcribe_audio(audio_path):
|
| 42 |
if not audio_path:
|
| 43 |
print("[LOG] Aucun fichier audio fourni.")
|
| 44 |
-
return "Aucun fichier audio fourni", [], None, []
|
| 45 |
|
| 46 |
print(f"[LOG] Début de la transcription de {audio_path}...")
|
| 47 |
result = pipe(audio_path, return_timestamps="word")
|
|
@@ -49,7 +42,7 @@ def transcribe_audio(audio_path):
|
|
| 49 |
|
| 50 |
if not words:
|
| 51 |
print("[LOG ERROR] Erreur : Aucun timestamp détecté.")
|
| 52 |
-
return "Erreur : Aucun timestamp détecté.", [], None, []
|
| 53 |
|
| 54 |
raw_transcription = " ".join([w["text"] for w in words])
|
| 55 |
word_timestamps = [(w["text"], w["timestamp"][0]) for w in words]
|
|
@@ -61,39 +54,31 @@ def transcribe_audio(audio_path):
|
|
| 61 |
return raw_transcription, [], audio_path, word_timestamps, transcription_with_timestamps
|
| 62 |
|
| 63 |
# -------------------------------------------------
|
| 64 |
-
# 3.
|
| 65 |
# -------------------------------------------------
|
| 66 |
-
def
|
| 67 |
-
print("[LOG]
|
| 68 |
formatted_data = []
|
| 69 |
for i, row in enumerate(table_data):
|
| 70 |
-
if
|
| 71 |
-
print(f"[LOG WARNING]
|
| 72 |
continue
|
| 73 |
|
| 74 |
-
text = row[0].strip()
|
| 75 |
segment_id = f"seg_{i+1:02d}"
|
| 76 |
-
|
| 77 |
-
|
| 78 |
-
|
| 79 |
-
for j, (word, start) in enumerate(word_timestamps):
|
| 80 |
-
if word in text.split():
|
| 81 |
-
if start_time is None:
|
| 82 |
-
start_time = start
|
| 83 |
-
end_time = word_timestamps[j+1][1] - 0.01 if j+1 < len(word_timestamps) else start + 0.5
|
| 84 |
-
|
| 85 |
-
formatted_data.append([text, start_time, end_time, segment_id])
|
| 86 |
-
print(f"[LOG] Segment ajouté : {text} | Début: {start_time}, Fin: {end_time}, ID: {segment_id}")
|
| 87 |
|
| 88 |
return formatted_data
|
| 89 |
|
| 90 |
# -------------------------------------------------
|
| 91 |
# 4. Validation et découpage des extraits audio
|
| 92 |
# -------------------------------------------------
|
| 93 |
-
def validate_segments(audio_path, table_data, metadata_state
|
| 94 |
print("[LOG] Début de la validation des segments...")
|
| 95 |
-
if not audio_path
|
| 96 |
-
print("[LOG ERROR] Erreur : Aucun
|
| 97 |
return [], metadata_state
|
| 98 |
|
| 99 |
if os.path.exists(TEMP_DIR):
|
|
@@ -104,27 +89,18 @@ def validate_segments(audio_path, table_data, metadata_state, word_timestamps):
|
|
| 104 |
segment_paths = []
|
| 105 |
updated_metadata = []
|
| 106 |
|
| 107 |
-
for
|
| 108 |
-
|
| 109 |
-
print(f"[LOG ERROR] Données incorrectes pour la validation : {row}")
|
| 110 |
-
continue
|
| 111 |
-
text, start_time, end_time, segment_id = row
|
| 112 |
-
|
| 113 |
-
if start_time is None or end_time is None:
|
| 114 |
-
print(f"[LOG ERROR] Timestamp manquant pour : {text}")
|
| 115 |
-
continue
|
| 116 |
-
|
| 117 |
-
start_ms, end_ms = int(float(start_time) * 1000), int(float(end_time) * 1000)
|
| 118 |
if start_ms < 0 or end_ms <= start_ms:
|
| 119 |
print(f"[LOG ERROR] Problème de découpage : {text} | {start_time}s - {end_time}s")
|
| 120 |
continue
|
| 121 |
|
| 122 |
segment_filename = f"{Path(audio_path).stem}_{segment_id}.wav"
|
| 123 |
segment_path = os.path.join(TEMP_DIR, segment_filename)
|
| 124 |
-
|
| 125 |
extract = original_audio[start_ms:end_ms]
|
| 126 |
extract.export(segment_path, format="wav")
|
| 127 |
-
|
| 128 |
segment_paths.append(segment_path)
|
| 129 |
updated_metadata.append({
|
| 130 |
"audio_file": segment_filename,
|
|
@@ -136,6 +112,7 @@ def validate_segments(audio_path, table_data, metadata_state, word_timestamps):
|
|
| 136 |
print(f"[LOG] Extrait généré : {segment_filename}")
|
| 137 |
|
| 138 |
return segment_paths, updated_metadata
|
|
|
|
| 139 |
# -------------------------------------------------
|
| 140 |
# 5. Génération du fichier ZIP
|
| 141 |
# -------------------------------------------------
|
|
@@ -163,6 +140,7 @@ def generate_zip(metadata_state):
|
|
| 163 |
|
| 164 |
print("[LOG] Fichier ZIP généré avec succès.")
|
| 165 |
return zip_path
|
|
|
|
| 166 |
# -------------------------------------------------
|
| 167 |
# 6. Interface utilisateur Gradio
|
| 168 |
# -------------------------------------------------
|
|
@@ -174,16 +152,16 @@ with gr.Blocks() as demo:
|
|
| 174 |
audio_input = gr.Audio(type="filepath", label="Fichier audio")
|
| 175 |
raw_transcription = gr.Textbox(label="Transcription", interactive=False)
|
| 176 |
transcription_timestamps = gr.Textbox(label="Transcription avec Timestamps", interactive=False)
|
| 177 |
-
table = gr.Dataframe(headers=["Texte"], datatype=["str"], row_count=(1, "dynamic"), col_count=
|
| 178 |
-
|
| 179 |
validate_button = gr.Button("Valider")
|
| 180 |
generate_button = gr.Button("Générer ZIP")
|
| 181 |
zip_file = gr.File(label="Télécharger le ZIP")
|
| 182 |
word_timestamps = gr.State()
|
| 183 |
|
| 184 |
audio_input.change(transcribe_audio, inputs=audio_input, outputs=[raw_transcription, table, audio_input, word_timestamps, transcription_timestamps])
|
| 185 |
-
|
| 186 |
-
validate_button.click(validate_segments, inputs=[audio_input, table, metadata_state
|
| 187 |
generate_button.click(generate_zip, inputs=metadata_state, outputs=zip_file)
|
| 188 |
|
| 189 |
demo.queue().launch()
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
import os
|
| 2 |
import shutil
|
| 3 |
import zipfile
|
| 4 |
import torch
|
|
|
|
|
|
|
| 5 |
import gradio as gr
|
| 6 |
+
from pathlib import Path
|
| 7 |
from pydub import AudioSegment
|
| 8 |
from transformers import pipeline
|
| 9 |
|
|
|
|
| 34 |
def transcribe_audio(audio_path):
|
| 35 |
if not audio_path:
|
| 36 |
print("[LOG] Aucun fichier audio fourni.")
|
| 37 |
+
return "Aucun fichier audio fourni", [], None, []
|
| 38 |
|
| 39 |
print(f"[LOG] Début de la transcription de {audio_path}...")
|
| 40 |
result = pipe(audio_path, return_timestamps="word")
|
|
|
|
| 42 |
|
| 43 |
if not words:
|
| 44 |
print("[LOG ERROR] Erreur : Aucun timestamp détecté.")
|
| 45 |
+
return "Erreur : Aucun timestamp détecté.", [], None, []
|
| 46 |
|
| 47 |
raw_transcription = " ".join([w["text"] for w in words])
|
| 48 |
word_timestamps = [(w["text"], w["timestamp"][0]) for w in words]
|
|
|
|
| 54 |
return raw_transcription, [], audio_path, word_timestamps, transcription_with_timestamps
|
| 55 |
|
| 56 |
# -------------------------------------------------
|
| 57 |
+
# 3. Enregistrement des segments définis par l'utilisateur
|
| 58 |
# -------------------------------------------------
|
| 59 |
+
def save_segments(table_data):
|
| 60 |
+
print("[LOG] Enregistrement des segments définis par l'utilisateur...")
|
| 61 |
formatted_data = []
|
| 62 |
for i, row in enumerate(table_data):
|
| 63 |
+
if len(row) < 3 or not row[0].strip():
|
| 64 |
+
print(f"[LOG WARNING] Ligne vide ignorée à l'index {i}.")
|
| 65 |
continue
|
| 66 |
|
| 67 |
+
text, start_time, end_time = row[0].strip(), row[1], row[2]
|
| 68 |
segment_id = f"seg_{i+1:02d}"
|
| 69 |
+
|
| 70 |
+
formatted_data.append([text, float(start_time), float(end_time), segment_id])
|
| 71 |
+
print(f"[LOG] Segment enregistré : {text} | Début: {start_time}, Fin: {end_time}, ID: {segment_id}")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 72 |
|
| 73 |
return formatted_data
|
| 74 |
|
| 75 |
# -------------------------------------------------
|
| 76 |
# 4. Validation et découpage des extraits audio
|
| 77 |
# -------------------------------------------------
|
| 78 |
+
def validate_segments(audio_path, table_data, metadata_state):
|
| 79 |
print("[LOG] Début de la validation des segments...")
|
| 80 |
+
if not audio_path:
|
| 81 |
+
print("[LOG ERROR] Erreur : Aucun fichier audio fourni !")
|
| 82 |
return [], metadata_state
|
| 83 |
|
| 84 |
if os.path.exists(TEMP_DIR):
|
|
|
|
| 89 |
segment_paths = []
|
| 90 |
updated_metadata = []
|
| 91 |
|
| 92 |
+
for text, start_time, end_time, segment_id in table_data:
|
| 93 |
+
start_ms, end_ms = int(start_time * 1000), int(end_time * 1000)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 94 |
if start_ms < 0 or end_ms <= start_ms:
|
| 95 |
print(f"[LOG ERROR] Problème de découpage : {text} | {start_time}s - {end_time}s")
|
| 96 |
continue
|
| 97 |
|
| 98 |
segment_filename = f"{Path(audio_path).stem}_{segment_id}.wav"
|
| 99 |
segment_path = os.path.join(TEMP_DIR, segment_filename)
|
| 100 |
+
|
| 101 |
extract = original_audio[start_ms:end_ms]
|
| 102 |
extract.export(segment_path, format="wav")
|
| 103 |
+
|
| 104 |
segment_paths.append(segment_path)
|
| 105 |
updated_metadata.append({
|
| 106 |
"audio_file": segment_filename,
|
|
|
|
| 112 |
print(f"[LOG] Extrait généré : {segment_filename}")
|
| 113 |
|
| 114 |
return segment_paths, updated_metadata
|
| 115 |
+
|
| 116 |
# -------------------------------------------------
|
| 117 |
# 5. Génération du fichier ZIP
|
| 118 |
# -------------------------------------------------
|
|
|
|
| 140 |
|
| 141 |
print("[LOG] Fichier ZIP généré avec succès.")
|
| 142 |
return zip_path
|
| 143 |
+
|
| 144 |
# -------------------------------------------------
|
| 145 |
# 6. Interface utilisateur Gradio
|
| 146 |
# -------------------------------------------------
|
|
|
|
| 152 |
audio_input = gr.Audio(type="filepath", label="Fichier audio")
|
| 153 |
raw_transcription = gr.Textbox(label="Transcription", interactive=False)
|
| 154 |
transcription_timestamps = gr.Textbox(label="Transcription avec Timestamps", interactive=False)
|
| 155 |
+
table = gr.Dataframe(headers=["Texte", "Début (s)", "Fin (s)"], datatype=["str", "number", "number"], row_count=(1, "dynamic"), col_count=3)
|
| 156 |
+
save_segments_button = gr.Button("Enregistrer les valeurs")
|
| 157 |
validate_button = gr.Button("Valider")
|
| 158 |
generate_button = gr.Button("Générer ZIP")
|
| 159 |
zip_file = gr.File(label="Télécharger le ZIP")
|
| 160 |
word_timestamps = gr.State()
|
| 161 |
|
| 162 |
audio_input.change(transcribe_audio, inputs=audio_input, outputs=[raw_transcription, table, audio_input, word_timestamps, transcription_timestamps])
|
| 163 |
+
save_segments_button.click(save_segments, inputs=table, outputs=table)
|
| 164 |
+
validate_button.click(validate_segments, inputs=[audio_input, table, metadata_state], outputs=[extracted_segments, metadata_state])
|
| 165 |
generate_button.click(generate_zip, inputs=metadata_state, outputs=zip_file)
|
| 166 |
|
| 167 |
demo.queue().launch()
|